museweb

Name: museweb
Author: kekePower

32stars1forks

How to install museweb?

kekepower/museweb

$git clone https://github.com/kekePower/museweb

Installs into the current project.

Summary

MuseWeb is an experimental, prompt-driven web server that streams HTML straight from plain-text prompts using a large-language model (LLM). Works with any OpenAI-compatible API - from local Ollama models to cloud providers like OpenAI, Anthropic, Google, Together.ai, Groq

Files · 1

View on GitHub

README.md

1	# MuseWeb
2
3	MuseWeb is an experimental, prompt-driven web server that streams HTML straight from plain-text prompts using a large-language model (LLM). Works with any OpenAI-compatible API - from local Ollama models to cloud providers like OpenAI, Anthropic, Google, Together.ai, Groq, and hundreds more. Originally built "just for fun," it currently serves as a proof-of-concept for what prompt-driven websites could become once local LLMs are fast and inexpensive. Even in this early state, it showcases the endless possibilities of minimal, fully self-hosted publishing.
4
5	Version 1.1.4 introduces enhanced model support, robust output sanitization, and critical streaming fixes for clean HTML generation.
6
7	---
8
9	## ✨ Features
10
11	* Prompt → Page – Point MuseWeb to a folder of `.txt` prompts; each prompt becomes a routable page.
12	* Live Reloading for Prompts – Edit your prompt files and see changes instantly without restarting the server.
13	* Streaming Responses – HTML is streamed token-by-token for instant first paint with real-time sanitization.
14	* Universal API Compatibility – Works with any OpenAI-compatible API endpoint:
15	* [Ollama](https://ollama.ai/) (default, runs everything locally)
16	* OpenAI (GPT-4, GPT-3.5, etc.)
17	* Anthropic Claude (via OpenAI-compatible proxies)
18	* Google Gemini (via OpenAI-compatible endpoints)
19	* Together.ai (hundreds of open-source models)
20	* Groq (ultra-fast inference)
21	* Inception Labs Mercury (advanced reasoning models)
22	* Perplexity (Sonar models with web search)
23	* Novita.ai (global model marketplace)
24	* OpenRouter (unified API for 200+ models)
25	* Local providers (LM Studio, vLLM, Text Generation WebUI, etc.)
26	* Any other OpenAI-compatible endpoint – Just change the `api_base` URL!
27	* Single Binary – Go-powered, ~7 MB static binary, no external runtime.
28	* Zero JS by Default – Only the streamed HTML from the model is served; you can add your own assets in `public/`.
29	* Modular Architecture – Clean separation of concerns with dedicated packages for configuration, server, models, and utilities.
30	* Prompt-Scoped Static Assets – Each prompt set can have its own `public/` directory for static files (CSS, images, JS, etc.), with automatic resolution and fallback to the global `public/` directory.
31	* Robust Output Sanitization – Advanced code fence removal and markdown artifact cleaning for pristine HTML output.
32	* Enhanced Model Support – Comprehensive support for reasoning models including DeepSeek, R1, Qwen, Mercury, and more.
33	* Configurable via `config.yaml` – Port, model, backend, prompt directory, and API credentials.
34	* Environment Variable Support – Falls back to `OPENAI_API_KEY` if not specified in config or flags.
35	* Reasoning Model Support – Automatic detection and handling of reasoning models with thinking output disabled for clean web pages.
36	* Detailed Logging – Comprehensive logging of prompt file loading and request handling for easy debugging.
37
38	---
39
40	## 🚀 Quick Start
41
42	```bash
43	# 1. Clone and build
44	$ git clone https://github.com/kekePower/museweb.git
45	$ cd museweb
46	$ GO111MODULE=on go build .
47
48	# 2. (Optional) pull an LLM with Ollama
49	$ ollama pull llama3
50
51	# 3. Run with defaults (localhost:8080)
52	$ ./museweb
53	```
54
55	Open <http://localhost:8080> in your browser. Navigation links are generated from the prompt filenames.
56
57	---
58
59	## 🔧 Configuration
60
61	Copy `config.example.yaml` to `config.yaml` and tweak as needed:
62
63	```yaml
64	server:
65	address: "127.0.0.1" # Interface to bind to (e.g., 127.0.0.1 or 0.0.0.0)
66	port: "8080" # Port for HTTP server
67	prompts_dir: "./prompts" # Folder containing *.txt prompt files
68	debug: false # Enable debug logging
69	model:
70	backend: "ollama" # "ollama" or "openai"
71	name: "llama3" # Model name to use
72	reasoning_models: # Patterns for reasoning models (thinking disabled automatically)
73	- "deepseek"
74	- "r1-1776"
75	- "qwen

People who liked this also viewed

6 picks

Type

museweb

32stars1forks

View on GitHub

How to install museweb?

kekepower/museweb

$git clone https://github.com/kekePower/museweb

Installs into the current project.

Summary

Files · 1

View on GitHub

README.md

1	# MuseWeb
2
3	MuseWeb is an experimental, prompt-driven web server that streams HTML straight from plain-text prompts using a large-language model (LLM). Works with any OpenAI-compatible API - from local Ollama models to cloud providers like OpenAI, Anthropic, Google, Together.ai, Groq, and hundreds more. Originally built "just for fun," it currently serves as a proof-of-concept for what prompt-driven websites could become once local LLMs are fast and inexpensive. Even in this early state, it showcases the endless possibilities of minimal, fully self-hosted publishing.
4
5	Version 1.1.4 introduces enhanced model support, robust output sanitization, and critical streaming fixes for clean HTML generation.
6
7	---
8
9	## ✨ Features
10
11	* Prompt → Page – Point MuseWeb to a folder of `.txt` prompts; each prompt becomes a routable page.
12	* Live Reloading for Prompts – Edit your prompt files and see changes instantly without restarting the server.
13	* Streaming Responses – HTML is streamed token-by-token for instant first paint with real-time sanitization.
14	* Universal API Compatibility – Works with any OpenAI-compatible API endpoint:
15	* [Ollama](https://ollama.ai/) (default, runs everything locally)
16	* OpenAI (GPT-4, GPT-3.5, etc.)
17	* Anthropic Claude (via OpenAI-compatible proxies)
18	* Google Gemini (via OpenAI-compatible endpoints)
19	* Together.ai (hundreds of open-source models)
20	* Groq (ultra-fast inference)
21	* Inception Labs Mercury (advanced reasoning models)
22	* Perplexity (Sonar models with web search)
23	* Novita.ai (global model marketplace)
24	* OpenRouter (unified API for 200+ models)
25	* Local providers (LM Studio, vLLM, Text Generation WebUI, etc.)
26	* Any other OpenAI-compatible endpoint – Just change the `api_base` URL!
27	* Single Binary – Go-powered, ~7 MB static binary, no external runtime.
28	* Zero JS by Default – Only the streamed HTML from the model is served; you can add your own assets in `public/`.
29	* Modular Architecture – Clean separation of concerns with dedicated packages for configuration, server, models, and utilities.
30	* Prompt-Scoped Static Assets – Each prompt set can have its own `public/` directory for static files (CSS, images, JS, etc.), with automatic resolution and fallback to the global `public/` directory.
31	* Robust Output Sanitization – Advanced code fence removal and markdown artifact cleaning for pristine HTML output.
32	* Enhanced Model Support – Comprehensive support for reasoning models including DeepSeek, R1, Qwen, Mercury, and more.
33	* Configurable via `config.yaml` – Port, model, backend, prompt directory, and API credentials.
34	* Environment Variable Support – Falls back to `OPENAI_API_KEY` if not specified in config or flags.
35	* Reasoning Model Support – Automatic detection and handling of reasoning models with thinking output disabled for clean web pages.
36	* Detailed Logging – Comprehensive logging of prompt file loading and request handling for easy debugging.
37
38	---
39
40	## 🚀 Quick Start
41
42	```bash
43	# 1. Clone and build
44	$ git clone https://github.com/kekePower/museweb.git
45	$ cd museweb
46	$ GO111MODULE=on go build .
47
48	# 2. (Optional) pull an LLM with Ollama
49	$ ollama pull llama3
50
51	# 3. Run with defaults (localhost:8080)
52	$ ./museweb
53	```
54
55	Open <http://localhost:8080> in your browser. Navigation links are generated from the prompt filenames.
56
57	---
58
59	## 🔧 Configuration
60
61	Copy `config.example.yaml` to `config.yaml` and tweak as needed:
62
63	```yaml
64	server:
65	address: "127.0.0.1" # Interface to bind to (e.g., 127.0.0.1 or 0.0.0.0)
66	port: "8080" # Port for HTTP server
67	prompts_dir: "./prompts" # Folder containing *.txt prompt files
68	debug: false # Enable debug logging
69	model:
70	backend: "ollama" # "ollama" or "openai"
71	name: "llama3" # Model name to use
72	reasoning_models: # Patterns for reasoning models (thinking disabled automatically)
73	- "deepseek"
74	- "r1-1776"
75	- "qwen

People who liked this also viewed

6 picks

Type